A Comprehensive Stress Test of Truncated Hilbert Space Bases against Green's function Monte Carlo in U(1) Lattice Gauge Theory
Timo Jakobs, Marco Garofalo, Tobias Hartung, Karl Jansen, Paul Ludwig, Johann Ostmeyer, Simone Romiti, Carsten Urbach
TL;DR
The paper tackles the challenge of digitising the infinite‑dimensional link Hilbert space in Hamiltonian lattice gauge theories by comparing several truncation schemes, including electric, magnetic, and a new plaquette‑state functional basis built from single‑plaquette eigenstates. It demonstrates that the plaquette‑state basis yields the most accurate results at a fixed operator dimension in 2D U(1) across a wide range of couplings and remains feasible in 3D, with Green’s function Monte Carlo providing cross‑checks where possible. Extending to SU(2) shows promising preliminary results for the plaquette basis, though a reliable local dual formulation for non‑Abelian theories remains an open challenge. The work supports the plaquette‑state approach as a compact, low‑truncation digitisation for tensor‑network and quantum computing simulations of LGTs and highlights GFMC as a valuable benchmarking tool for Hamiltonian methods.
Abstract
A representation of Lattice Gauge Theories (LGT) suitable for simulations with tensor network state methods or with quantum computers requires a truncation of the Hilbert space to a finite dimensional approximation. In particular for U(1) LGTs, several such truncation schemes are known, which we compare with each other using tensor network states. We show that a functional basis obtained from single plaquette Hamiltonians -- which we call plaquette state basis -- outperforms the other schemes in two spatial dimensions for plaquette, ground state energy and mass gap, as it is delivering accurate results for a wide range of coupling strengths with a minimal number of basis states. We also show that this functional basis can be efficiently used in three spatial dimensions. Green's function Monte Carlo appears to be a highly useful tool to verify tensor network states results, which deserves further investigation in the future.
